Premium
Trust separation on the Cray XC40 using PBS Pro
Author(s) -
Clarke Sam
Publication year - 2017
Publication title -
concurrency and computation: practice and experience
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.309
H-Index - 67
eISSN - 1532-0634
pISSN - 1532-0626
DOI - 10.1002/cpe.4296
Subject(s) - supercomputer , scalability , partition (number theory) , lustre (file system) , computer science , set (abstract data type) , database , operating system , mathematics , combinatorics , programming language
Summary As the UK's national weather agency, the Met Office has a requirement to produce regular, timely weather forecasts. As a major centre for climate and weather research, it has a need to provide access to large‐scale supercomputing resources to users from within the organisation. It also provides a supercomputer facility for academic partners inside the UK, and to international collaborators. Each of these user categories has a different set of availability requirements and requires a different level of access. This paper describes the steps taken to create an HPC facility that separates these different requirements using soft partitions created by the batch system. We detail our initial experiences with cgroup containers and our use of custom PBS hooks to partition the Lustre name space. We summarise some of the problems observed during implementation, comment on the scalability of the solution and outline possible future enhancements.